Path Planning Framework for Unmanned Ground Vehicles on Uneven Terrain

In this thesis, I address the problem of long-range path planning on uneven terrain for non-holonomic wheeled mobile robots (WMR). Uneven terrain path planning is essential for search-and-rescue, surveillance, military, humanitarian, agricultural, constructing missions, etc. These missions necessitate the generation of a feasible sequence of waypoints, or reference states, to navigate a WMR from the initial location to the final target location through the uneven terrain. The feasibility of navigating through a given path over uneven terrain can be undermined by various terrain features. Examples of such features are loose soil, vegetation, boulders, steeply sloped terrain, or a combination of all of these elements. I propose a three-stage framework to solve the problem of rapid long-range path planning. In the first stage, RRT-Connect provides a rapid discovery of the feasible solution. Afterward, Informed RRT* improves the feasible solution. Finally, Shortcut heuristics improves the solution locally. To improve the computational speed of path planning algorithms, we developed an accelerated version of the traversability estimation on point clouds based on Principal Component Analysis. The benchmarks demonstrate the efficacy of the path planning approach.